Hacker News Reader

Comment by electrograv | Hacker News Reader

> Netflix said in a statement that “the reviews and our redesigned ratings system (thumbs up/down) never contributed to how we approach personalizing recommendations for members and writing a ‘bad’ review never had any bearing to whether a title was recommended to another viewer or not.”

THAT explains why Netflix keeps recommending me garbage quality content! I had a suspicion this was true, but couldn’t imagine anyone would make such a terrible design choice: 95% of content Netflix recommends me now is unwatchably bad in content quality (including Netflix originals too, of course).

IMO abolishing user ratings is quite probably the worst possible solution to losing user engagement due to low quality content.

View on HN

How in the world does this make sense? The entire competitive advantage of having the scale and user information that Netflix has is to be able to do personal recommendations by recommending shows based on what others with similar taste like. Heck, I know very little about big data and even I have walked through a simple sentiment analysis proof of concept demo to rate comments as positive or negative.

They are probably determining this directly on when someone is watching a show, which shows, etc.

I can only watch shows they recommend though! You get stuck in a local maxima with no way to escape. God help you if you watch a WWII documentary.

I so rarely find a title I search for that I don't bother with that anymore.

> I can only watch shows they recommend though!

That's not true at all. You can absolutely browse by genre, actor, etc.

not on my smart tv (tizen), or at least not in any sensible way.

It's still heavily colored by your history. Using someone else's Netflix was really an eye opener. This new algo driven world is stifling.

Sometimes I search for a genre, minor actor, or just a random word. I almost always find something I'd like to watch.

Really? Because today, August 19th 2018, I search for Sci-Fi films and get like 20 totals results. Moon is not even in the Sci-Fi Genres for f*s sake. None of the greaty Sci-Fi indie films I used to get recommended to me 4 years ago even show up in a Sci-Fi search. Netflix has truly succumbed to their hiring/firing practices IMHO.

Did you browse the sci-fi directory or did you do a word search on "Sci-Fi." I went to the directory and Moon was the first title in the list.

Why take someone's word for what they like or don't like when you know what they actually watch?

How do you know a user liked a movie or TV show after watching it? I think a movie has to be really really bad for people to actually stop halfway through. For TV shows it might be more likely that people drop out after a couple episodes but if it's just "meh" or the show's finale was disappointing, how do your algorithms know?

Once you can analyze all the things someone watches and all the views on a single video, then it's not particularly useful to know if this one person liked this one video.

I'm not sure how much it matters how good each episode of a tv show is. Do you look up the rating of each subsequent episode before you watch it? If most people in your cohort were hooked enough for 5 episodes before tapering off, then certainly that's a tv show you may want to see.

Also, if everyone was interested enough to sit through something even though most people would say the payoff sucked, as opposed to bailing on the video early on, then I'm not convinced it's much different from a movie that people sat through and particularly liked when it comes to recommendation. So people enjoyed 75/90min instead the full 90/90min, not very damning.

They won't watch the next show in the same category or series. Or the next. Seeing one 'meh' to the end is one thing, seeing two to the end is unlikely, seeing three and they are lying in their comments. I fact I believe minutes watched are more accurate than ratings. Very few people take time to rate, and of that category even less rate properly for algorithms to make sense.

Still, they could have left it in and just collect the data. Even if for providing a false sense/feeling of choice and ownership in the matter ;).

They not only know what you watch, but know which parts you re-watch.

When I find a TV show "meh", I usually jump from episode 3-4 to season finale to get closure. I wonder if others do this.

That's how we got the hate read content economy.

Imagine Amazon doing the same.

If I am going to pay for something on Amazon sight unseen, I care about reviews. But I’ve already paid for Netflix. There is no additional monetary commitment.

There is, it's the next month and the next month and so on and so forth.

It got bad enough for me that I deleted my Netflix account a few months ago and haven't really missed it. I got tired of scrolling through for ages and never finding anything I felt like watching. It still surprises me that they don't do a better job of recommendations given they used to make a big deal out of it.

So, what are you watching instead?..

Well since then my TV broke and I haven't bothered replacing it, so nothing. We haven't really missed it but that might change when summer is over, we'll see.

Youtube has a good recommendation engine (mostly based on last 10 videos you viewed), so it is enjoyable to browse Youtube videos.

A long time ago netflix would allow you to go into a genre and sort movies by star rating. Finding something good to watch was very straight forward back then.

Have you actually tried to watch what Netflix recommends to you or do you just feel like it's bad? I'm definitely in the latter case, but I haven't been trying to prove it.

Reply on news.ycombinator.com